Algorithms of Feature Extraction Methods

878 Words2 Pages

Algorithms of feature extraction methods

1. Intersection points

Principle:
• Number and position of intersections between the character and straight lines have been retained as features.
• 1st horizontal (1/3)rd of height.;
• 2nd horizontal (2/3)rd of height.;
• The vertical straight line goes through the center of gravity of the character

Algorithm:
INPUT: Character image
OUTPUT : Number and position of intersections between the character and straight lines begin Step 1: Transform the input image into thinned image
Step 2: Draw two horizontal straight lines at the first third and the second third of the height of the character
Step 3: The vertical straight line goes through the centre of gravity of the character
Step 4: Find the X-position of the intersection point of horizontal line at1/3rd of height of character with the character.
Step 5: Find the X-position of the intersection point of horizontal line at 2/3rd of height of character with the character.
Step 6: Find the Y-position of the intersection point of vertical line at centre of gravity of character with the character.
Step 7: Store above features to a file. end • No. of horizontal intersection points 2
• No. of vertical intersection points 1
• Total no. of Intersection Points 3

2. Zone Centroid Features

Principle:
• The major advantage of this approach stems from its robustness to small variation, ease of implementation and good recognition rate.
• Zone-based feature extraction method provides good result even when certain pre processing steps like filtering, smoothing and slant removing are not considered.
Algorithm 1: Image Centroid and Zone-based (ICZ) Distance Metric Feature Extrac...

... middle of paper ...

...orizontal Profile =( left side view )/(right side view )

3) Downside View: the sum of white pixels when scanning an input image vertically from the down (y2) to the up (y3) until it touches a black pixel.

3) Upside View: the sum of white pixels when scanning an input image vertically from the up (y0 ) to the down (y1) until it touches a black pixel.

 Vertical profile= (upside view)/(downside view)

Input Image Size (128 X 128) left= 0 right= 582
HP= 0

Image size (256 X 256) left= 0 right= 2223
HP= 0

The characters B, D, E, F, K, L, P, R has left profile value as 0
The Horizontal profile values are scale invariant

8. Zone Density (ZD) Features

• density of each zone is obtained by dividing the number of foreground pixels in each zone by total number of pixels in each zone.

16 zones of 32*32 normalized Devnagari numeral „3‟.

More about Algorithms of Feature Extraction Methods

Open Document